Qwen-Image 2.1 open-weights 7B model gets Day-0 ComfyUI support, runs in 30s on a laptop
Lexius2129 · reddit · 2026-09-21
ComfyUI shipped Day-0 support for Qwen-Image 2.1, an open-weights (research license) 7B unified generation and editing model based on the Qwen-Image 2.0 architecture. It supports up to 10 input images, strong text rendering, native 2K output, and alpha/transparency generation. On a 5080 laptop (16GB VRAM) the int8 workflow averages 36s with the bf16 text encoder, 42s with Kijai's w4a8 quantized encoder, and just 30s when combined with the comfy kitchen attention backend. Weights and workflow templates are on Hugging Face and GitHub.
More from coding & agent
- Hermes Agent Now Streams Its Reasoning Live Into Any OpenAI-Compatible App — Teknium · 2026-09-21
- Devs clash over whether beginners should still learn to code as AI automates code review — mgill25 · 2026-09-21
- Pair vibecoding with a novice reveals the real unlock: communication and systems thinking — eschadiol · 2026-09-21
- Substrate bets the agent loop is the new scheduling loop with fast suspend/resume — rakyll · 2026-09-21
- Long-running coding agents let you delegate work all weekend — josh_bickett · 2026-09-21
- Many Mac mini buyers would pay ~$1k for always-on cloud Codex/Claude Code, says alliekmiller — alliekmiller · 2026-09-21